Improved Hybrid Memory Cube for Weight-Sharing Deep Convolutional Neural Networks | IEEE Conference Publication | IEEE Xplore